US10359339B2 - Monitoring system for an engine test bench - Google Patents

Monitoring system for an engine test bench Download PDF

Info

Publication number
US10359339B2
US10359339B2 US13/878,107 US201113878107A US10359339B2 US 10359339 B2 US10359339 B2 US 10359339B2 US 201113878107 A US201113878107 A US 201113878107A US 10359339 B2 US10359339 B2 US 10359339B2
Authority
US
United States
Prior art keywords
exogenic
test bench
indicator vector
engine component
indicator
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active, expires
Application number
US13/878,107
Other languages
English (en)
Other versions
US20130211768A1 (en
Inventor
Valerio Gerez
Jerome Henri Noel Lacaille
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Safran Aircraft Engines SAS
Original Assignee
Safran Aircraft Engines SAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Safran Aircraft Engines SAS filed Critical Safran Aircraft Engines SAS
Assigned to SNECMA reassignment SNECMA ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: GEREZ, VALERIO, LACAILLE, JEROME HENRI NOEL
Publication of US20130211768A1 publication Critical patent/US20130211768A1/en
Assigned to SAFRAN AIRCRAFT ENGINES reassignment SAFRAN AIRCRAFT ENGINES CHANGE OF NAME (SEE DOCUMENT FOR DETAILS). Assignors: SNECMA
Assigned to SAFRAN AIRCRAFT ENGINES reassignment SAFRAN AIRCRAFT ENGINES CORRECTIVE ASSIGNMENT TO CORRECT THE COVER SHEET TO REMOVE APPLICATION NOS. 10250419, 10786507, 10786409, 12416418, 12531115, 12996294, 12094637 12416422 PREVIOUSLY RECORDED ON REEL 046479 FRAME 0807. ASSIGNOR(S) HEREBY CONFIRMS THE CHANGE OF NAME. Assignors: SNECMA
Application granted granted Critical
Publication of US10359339B2 publication Critical patent/US10359339B2/en
Active legal-status Critical Current
Adjusted expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M15/00Testing of engines
    • G01M15/02Details or accessories of testing apparatus
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M15/00Testing of engines
    • G01M15/04Testing internal-combustion engines
    • G01M15/05Testing internal-combustion engines by combined monitoring of two or more different engine parameters

Definitions

  • This invention relates to the domain of engine test bench monitoring systems.
  • the engine may be of any type and it may be designed for installation in land, sea or aerospatial vehicles.
  • the information disclosed in this invention is applicable to an aircraft engine, although this example is not limitative.
  • Engines use very sophisticated technologies and are systematically tested on test benches. During a test on a test bench, the engine or engine components and the bench are monitored using a large number of sensors. Some tests in the engine development phase may require recording of more than a thousand measurements including performance measurements at low frequency up to 100 Hz (for example pressures, temperatures, gauges, etc.), dynamic measurements at high frequency up to 50 kHz (for example measurements output from accelerometers, microphones, etc.) and some measurements or data describing the test procedure.
  • performance measurements at low frequency up to 100 Hz for example pressures, temperatures, gauges, etc.
  • dynamic measurements at high frequency up to 50 kHz for example measurements output from accelerometers, microphones, etc.
  • SPC Statistic Process Control
  • this SPC testing system simply collects monitored measurements and verifies that they remain between two safety thresholds. Since variations in the measurement context are not taken into account, the safety thresholds of the SPC test system must necessarily be very large. Thus, it is very difficult for such a system to detect an operating error and it simply protects the test bench by triggering warnings in the case of a very serious failure.
  • test bench operators usually try to test the engine under extreme operating conditions. This can create safety problems and serious risks of damaging the bench which is very expensive, which can cause engine development delays.
  • the purpose of this invention is to disclose a system for monitoring a test bench capable of precisely and safely monitoring operation of the test bench and the engine or component of the engine under test.
  • This invention is defined by a monitoring system for a test bench for at least one engine component comprising:
  • this monitoring system is a self-adaptive system capable of managing a changing and multi-contextual environment.
  • the system adapts to changing configurations by automatically detecting context changes so as to enable the anomaly detector to make a local diagnostic of measurements to be monitored as a function of the identified context.
  • an anomaly detector on a test bench also helps to validate it before it is used, for example on an onboard engine.
  • the monitoring system comprises:
  • the system learns to perform an automatic and unsupervised classification of contexts in order to identify the different operating modes of the test bench and engine component combination. This enables the monitoring system to manage all random manoeuvres that may be applied to the engine under test or an engine component under test.
  • the processing means are configured to update said set of context classes by checking if a new detection of an exogenic indicator vector belongs to a previously constructed context class, and recording the exogenic indicator vector of the new detection in a database if there is no context class that corresponds to the new detection until an appropriate number of similar exogenic indicator vectors have been detected to form a new context class.
  • the system continuously relearns to improve the unsupervised classification of contexts in order to achieve a stable and fairly robust classification after a certain time.
  • processing means are configured to update said set of context classes verifying if new detections of exogenic indicator vectors belong to previously constructed context classes and recording at least some of said new detections in the corresponding context classes.
  • this makes it possible to regularly make a new update of each class in order to take account of slow changes to operating modes in order to further improve monitoring of the test bench and engine component combination.
  • the processing means are configured to identify the context class of an exogenic indicator vector by calculating a match value of said exogenic indicator vector relative to each context class.
  • the match value is used to verify whether new detections of exogenic indicator vectors resemble vectors that have already been used for learning.
  • the match value is a credible and quantifiable indication about whether an exogenic indicator vector belongs to a context class and which quickly becomes more reliable over time. Consequently, when the match value becomes high enough, the monitoring system can begin to make precise and relevant diagnostics.
  • the processing means are configured to construct said set of context classes using a likelihood maximisation criterion.
  • This iterative criterion can be used to fairly quickly obtain a precise context dependent classification.
  • the processing means are configured to select an appropriate number of context classes based on optimisation criterion applied to the exogenic indicator vectors.
  • the monitoring system comprises:
  • test bench and engine component combination can be reliably transformed into indicator vectors that take account of local and global variations of measurements, and thus producing an unsupervised, adaptative and precise classification of operating modes.
  • the processing means are configured to calculate a quality value of said risk probability.
  • the anomaly detector implements a normal behaviour model and generates a normality measurement (score) by a likelihood calculation.
  • the anomaly detector implements a bearing damage detection model.
  • This second detector can work offline, possibly regularly, to produce a diagnostic of wear and damage of test bench and engine bearings.
  • the anomaly detector implements an intermittent events detection model.
  • This third detector is capable of detecting the occurrence of intermittent vibration events in the test bench and the engine.
  • the monitoring system comprises supervision means in which said anomaly detectors are encapsulated, said supervision means being configured to start said anomaly detectors, to procure input data for each of said anomaly detectors, to receive output messages from each of said anomaly detectors and to manage instances corresponding to parameter settings and calibration choices for each of said anomaly detectors.
  • the invention also relates to a management system comprising a control system connected to an engine test bench, the control system being configured to control the test bench and to record data output from the test bench and at least one component of an engine in a storage means, the management system also comprising a monitoring system according to any one of the previous characteristics, said monitoring system being connected to the test bench through the control system that sends said data to it output from the test bench and engine component combination.
  • control system can be deployed in a first computer and the monitoring system can be deployed in a second computer at a distance from the first.
  • the invention also relates to a method of monitoring a test bench for at least one engine component, comprising the following steps:
  • the invention is also related to a computer program comprising code instructions for implementation of the method according to the claim when it is executed by a computer.
  • FIG. 1 diagrammatically shows a monitoring system for a test bench for at least one engine component according to the invention
  • FIG. 2 is a flow chart according to one particular embodiment of the invention showing steps in the construction of an endogenic indicator vector and an associated exogenic indicator vector;
  • FIG. 3 is a flow chart according to one particular embodiment of the invention showing steps in the construction of a set of context classes
  • FIG. 4 is a flow chart showing steps to select an optimum number of context classes according to the invention.
  • FIG. 5 is a flow chart showing an example encapsulation of anomaly detectors in the monitoring system according to the invention.
  • FIG. 6 diagrammatically shows encapsulation of anomaly detectors in the monitoring system according to the invention.
  • FIG. 7 diagrammatically shows a management system for the test bench for an engine component according to the invention.
  • the basic concept of the invention is to propose a test bench monitoring system using monitoring mechanisms designed for onboard aircraft engines. These monitoring mechanisms (onboard) are configured to detect anomalies, knowing engine operating modes in advance.
  • an onboard engine always operates in the same way following clearly defined flight phases, as follows: start engine, taxi, take off, climb, cruise, approach, landing, reverse and stop engine.
  • start engine taxi, take off, climb, cruise, approach, landing, reverse and stop engine.
  • engine operating modes by fairly simple indicators based on elementary criteria such as engine speed, altitude and aircraft attitudes.
  • engine component will be used throughout the remainder of this description to denote the entire engine or simply an engine component.
  • the manoeuvres applied to an engine component installed on a test bench may be very changing, unpredictable and may include extreme conditions. Furthermore, there is often a need to test new equipment. Thus, it is impossible to predict or classify operating modes of the test bench and the engine component in advance and consequently, normal onboard aircraft engine monitoring mechanisms cannot operate in such an environment.
  • one purpose of this invention is to make an unsupervised classification and an automatic identification of contexts in an environment consisting of a test bench and an engine component.
  • FIG. 1 diagrammatically shows a monitoring system 1 for a test bench 3 of at least one component 5 of an aircraft engine, according to the invention.
  • This system 1 comprises data acquisition means 7 and information processing means 9 such as a computer to execute one or several computer programs including program code instructions stored in storage means 10 of the computer and designed to use the monitoring system of the combination 11 of the test bench 3 and the engine component 5 .
  • the engine component 5 is mounted in the test bench 3 and a large number of sensors 13 (a few hundred) are used to acquire measurements on the test bench 3 and on the engine component 5 . These measurements are retrieved in the form of digital channels through acquisition buses or channels 15 at constant frequencies and are transmitted to the monitoring system 1 . Some measurements (for example vibration measurements) are recorded at high frequency (of the order of 50 kHz) while other measurements (for example pressure measurements) are recorded at low frequency (about 1 Hz to 100 Hz).
  • the set of measurements can be subdivided into subsets of measurements related to different elements of the combination 11 of the test bench 3 and the engine component 5 based on criteria produced by expertise.
  • the shaft line that corresponds to mechanical coupling between the test bench 13 and the engine 5 knowing that the engine is driven in rotation by a shaft that is turned by the bench 3 .
  • the subset of measurements relative to the shaft line is selected, possibly comprising low frequency measurements (for example rotation speeds, pressures, temperatures, etc.) and high frequency measurements (for example accelerations, displacements).
  • an exogenic parameter is a context dependent parameter that is representative of the context (in other words the operating mode or functional conditions) of the combination 11 of the test bench 3 and the engine component 5 .
  • an endogenic parameter is a parameter that can be observed and is to be monitored and analysed as a function of its observation context to detect an anomaly. Endogenic parameter and exogenic parameters may be identified based on criteria produced by expertise.
  • exogenic parameter measurements comprise rotation speeds, inlet air pressures, temperatures, etc.
  • endogenic parameter measurements comprise vibration measurements, energies, shaft displacements, unbalanced masses, etc. Obviously information about energy or unbalanced mass is quite different in different contexts.
  • the monitoring system 1 is configured to transform flows of parameter measurement data from combination 11 of the test bench 3 and the engine component 5 into indicators to identify contexts automatically and to detect anomalies as a function of the context.
  • acquisition means 7 are configured to acquire time signal packets corresponding to measurements of endogenic and exogenic parameters specific to the combination 11 of the test bench 3 and the engine component 5 at successive instants (for example at regular instants).
  • the processing means 9 are configured to construct an endogenic indicator vector and an associated exogenic indicator vector starting from packets of time signals earlier than the current instant.
  • An endogenic or exogenic indicator vector may be constructed by compressing time signal packets as described below with reference to FIG. 2 .
  • each indicator vector (endogenic or exogenic) may for example be calculated periodically from past data.
  • processing means 9 are configured to identify a context class for the exogenic indicator vector constructed at the current instant.
  • a context class may be identified automatically for example by calculating distances of an exogenic indicator vector at the current instant relative to exogenic indicator vectors constructed at previous instants by defining the class by a determined number of the closest exogenic vectors.
  • a context class may also be identified automatically without supervision but by analysing a class among a set of classes constructed by learning to which it belongs, as described below with reference to FIG. 3 .
  • the processing means 9 can use at least one anomaly detector (for example designed for use on an onboard aircraft engine) to calculate a score or risk probability of the current endogenic indicator vector conditioned by the context class identified for the associated exogenic indicator vector in order to make a diagnostic of the state of the combination 11 of the test bench 3 and the engine component 5 .
  • at least one anomaly detector for example designed for use on an onboard aircraft engine
  • the processing means 9 are also configured to calculate a quality or precision value of the risk probability that can help to evaluate the relevance of the risk.
  • FIG. 2 is a flow chart according to one particular embodiment showing steps in the construction of an endogenic indicator vector and an associated exogenic indicator vector.
  • step E 1 the processing means 9 analyse time signal packets corresponding to measurements of parameters A, B, . . . , G, H received at successive instants and identify measurements of endogenic parameters A, B, etc. and exogenic parameters G, H, etc.
  • the endogenic parameter measurements are used to construct endogenic indicator vectors and measurements of exogenic parameters are used to construct exogenic indicator vectors.
  • the term “parameter” is used in the remainder of this description to denote either an “endogenic parameter” or an “exogenic parameter” indifferently, and similarly the term “indicator vector” is used to denote an “endogenic indicator vector” or an “exogenic indicator vector” indifferently.
  • step E 2 the time signal packets of the different parameter measurements are recorded in buffer memories 21 a - 21 h at defined frequencies. More particularly, the processing means 9 are configured to buffer at least one packet of time signals earlier than the current instant in the variable size buffer memories 21 , for each parameter.
  • At least one time interval is defined to record the corresponding time signal.
  • the last ten seconds of the corresponding time signal may be recorded in a first buffer memory 21 a and the last twenty seconds may be recorded in a second buffer memory 21 b , etc.
  • several buffer memories with different and/or the same sizes can be defined for each parameter.
  • a first time signal A 1 and a second time signal A 2 of parameter A are recorded in first and second buffer memories 21 a , 21 b , etc.
  • time signals B 1 , B 2 , . . . , G 1 , G 2 , H 1 , H 2 of parameters B, . . . , G, H are recorded in buffer memories 21 c , 21 d , . . . , 21 h respectively.
  • several packets of corresponding signals can be buffered in several buffer memories 21 a - 21 h , for each parameter.
  • step E 3 the processing means 9 are configured to smooth each packet of time signals at at least one scale to form curves A 11 , A 21 , . . . , H 11 , H 12 representative of these packets.
  • Smoothing is a convolution operation that creates a global or local representation of time signal packets at the chosen scale. Strong smoothing illustrates the variation or the global tendency of the signals, while weak smoothing determines the local behaviour of the signals.
  • processing means are configured to re-sample the representative curves A 11 , A 21 , . . . , H 11 , H 12 formed in the previous step. Once a smoothed curve is available, a few representative points corresponding to a first compression may be sufficient. Thus, at the end of this step, the time signal packets of each parameter A, . . . H are transformed into a sequence of re-sampled curves a 11 , a 21 , . . . , h 21 that are small curves of a few points (for example of the order of about ten points).
  • the re-sampled curve Y t (in other words a 11 , or a 21 , . . .
  • Y t [y t , y t-r , . . . , y t-(n-1)r) ] where y t is the convolution of x t by the filter a according to the following formula:
  • step E 5 the processing means 9 are configured to compress the re-sampled curves a 11 , a 21 , . . . , h 21 in order to construct the endogenic or exogenic indicator vector. Compression is done individually for each re-sampled curve Y t (in other words for each of the small curves a 11 , a 21 , . . . , h 21 ).
  • each re-sampled curve Y t can be compressed using a main component analysis.
  • a basic change is then made by projecting each normalised curve Y t on an orthonormal base of curves-models (or curve-templates) ⁇ 1 , ⁇ 2 , . . . , ⁇ n ⁇ , each curve-model ⁇ t being a basic vector with size n.
  • Each re-sampled curve Y t may then be expressed by a series of curves-models as follows:
  • the values ⁇ t,1 , ⁇ t,2 , . . . , ⁇ t,k are projection coefficients of, the normalised curve Y t on the orthonormal base and ⁇ t is the residual error of the compression when only the k first curves-models ⁇ l to ⁇ k (k ⁇ 0) are used.
  • the main component analysis classifies the curves-models in order of importance and eliminates those that are not important or more precisely, groups together those for which there is a small variance in the remainder ⁇ t . As a result, the optimum set of curve-model ⁇ i with size k is obtained that minimises the norm of the remainder ⁇ t ⁇ .
  • FIG. 3 is a flow chart according to a particular embodiment showing steps in the construction of a set of context classes.
  • the processing means 9 are configured to construct an initial set of context class C 1 , C 2 , . . . , CK 0 starting from a learning sequence of initial exogenic indicator vectors ⁇ t1 , ⁇ t2 , . . . , ⁇ tN ⁇ . This construction is done without supervision, in other words, without any advanced knowledge of class names.
  • the processing means 9 are then configured in an execution phase (steps E 22 -E 26 ) to update this set of context classes self-adaptively starting from new inputs of exogenic indicator vectors.
  • step E 22 the processing means 9 are configured to verify if a new detection of an exogenic indicator vector ⁇ n belongs to a previously constructed context class. For example, this verification may be made by calculating a match value of the exogenic indicator vector ⁇ n relative to each of the context classes and identifying the class that gives the best match.
  • the match concept is described in the publication by J. Lacaille “ Validation of health monitoring algorithms for civil aircraft engines, IEEE Aerospace Conference, Big Sky, Mont., 2010”.
  • the match value may be considered as being the measurement of a distance between the exogenic indicator vector and a context class, and thus can be used to verify if the exogenic indicator vector of the new detection resembles vectors that have already been used for learning. If the test in step E 22 is positive, then the operation proceeds to step E 23 , otherwise step E 25 will be started.
  • Step E 23 concerns identification of the context class. If the exogenic indicator vector ⁇ n of the new detection belongs to a previously constructed context class, then this belonging identifies the context class and is used to make diagnostics in step E 24 using an anomaly detector.
  • at least some of these new detections of exogenic indicator vectors are recorded in the database 10 so that they can then be recorded in the corresponding context classes.
  • the exogenic indicator vector ⁇ n of the new detection is recorded in the database 10 in step E 25 , until an appropriate number of similar exogenic indicator vectors have been detected to form a new context class.
  • step E 26 the exogenic indicator vectors recorded in the database 10 are used to update the context classes C 1 , C 2 , . . . , CK and to relearn the classification when the match between the exogenic indicator vectors and the existing classes is weak.
  • the set of context classes in steps E 21 to E 26 can be constructed using a maximisation likelihood criterion.
  • a maximisation likelihood criterion it is assumed that the exogenic indicator vector ⁇ follows a normal distribution inside each context class, and an EM (Expectation-Maximisation) type maximisation criterion can be used that consists of identifying a mix of Gaussian densities (for example see the document by Dempster et al., “ Maximum likelihood from incomplete data via the EM algorithm Journal of the Royal Statistical society, 39(1):1-38, 1977)”.
  • the EM method is an iterative process that converges towards a model of coefficients for Gaussian class identifier laws.
  • a class identifier is used to calculate the match between an exogenic indicator vector and the class, and therefore to determine whether or not an exogenic indicator vector belongs to an existing class (see steps E 23 and E 25 ).
  • the EM method also needs to be initialised with an initial number of classes.
  • FIG. 4 is a flow chart showing steps for selecting an optimum number of classes in the context of an EM type maximisation criterion.
  • step E 31 an initial number K 0 of context classes is fixed.
  • step E 32 a classification mechanism is used (for example EM) to construct context classes from detections of exogenic indicator vectors as shown in FIG. 3 .
  • an optimisation criterion is applied on exogenic indicator vectors to select the appropriate number of context classes.
  • a likelihood coefficient is calculated, for example based on a likelihood criterion or a BIC (Bayesian Information Criterion) type criterion.
  • the method then consists of minimising the BIC coefficient. This coefficient reduces when the likelihood increases, but is penalised by the number K of classes.
  • step E 34 the number of context classes is increased (step E 34 ) and the previous steps are restarted, by looping back to step E 32 .
  • the likelihood coefficient is minimal, then learning is stopped and then a fairly robust classification is obtained in step E 35 .
  • This coefficient can thus be used to determine the best number of classes to reach an optimum classification.
  • the number K of classes is of the order of ten and the number N of detections is of the order of a hundred.
  • a robust classification can create a stochastic contexts law model and consequently create a high precision diagnostic of the state of the test bench and the engine component.
  • U be the random variable of exogenic indicator vectors
  • X the random variable of endogenic indicator vectors
  • An example of an anomaly detector for an aircraft engine is described in the publication by J. Lacaille “ Standardized failure signature for a turbofan engine IEEE Aerospace Conference, Big Sky, Mont., 2009”.
  • This anomaly detector is composed of two blocks.
  • the first block standardises and normalises the indicators eliminating local dependences relative to the acquisition context while managing stochastic interdependence relations between the indicators themselves.
  • the second block consists of modelling indicators and calculating an anomaly score or a risk probability R(t) starting from the likelihood of the model.
  • the score may be obtained from the remainder of a generalised linear regression.
  • an estimator of the conditional variance of the model is used to define the precision Pr(t) or the reliability of the calculated risk.
  • An anomaly may be confirmed after several successive detections and possibly after corroboration by various anomaly sources. If the anomaly is confirmed, the monitoring system 1 trips or issues a warning message.
  • anomaly score may also be associated with a class number and possibly a label. Labelling of anomalies and damage from anomalies within each class can be defined by expertise.
  • FIG. 5 is a flow chart showing an example encapsulation of anomaly detectors in the monitoring system 1 .
  • an anomaly detector is used for each context class C 1 , . . . , CK.
  • the input data to step E 41 consists of segments of first time data to be monitored (in other words time signal packets corresponding to endogenic parameter measurements) and a second input (in other words time signal packets corresponding to exogenic parameter measurements) is associated with each first input data describing context-sensitive aspects of the functioning in progress.
  • the first and second input data at a current instant t are transformed into exogenic and endogenic indicator vectors ( ⁇ t , ⁇ tilde over (x) ⁇ t ).
  • the exogenic indicator vector is used to identify the context class or functioning mode as described above.
  • step E 42 a finer compression 41 - 43 is made on the time data to be monitored that are specific to each identified context class. In this case, compression is fairly easy because data belong to the same class and are more similar.
  • an anomaly detector 51 - 53 is used for each context class C 1 , . . . , CK.
  • Each generic anomaly detector 51 - 53 produces scalar information by a data table (between 0 and 1) that represents the risk of anomaly. This output is associated with the quality value giving an estimate of the relevance of the result (between 0 and 1).
  • a warning message is issued in step E 44 if the anomaly is confirmed. For example, a warning is triggered if a first threshold is exceeded and this overrun is confirmed. Furthermore, an anomaly prognostic warning may be triggered if it is anticipated that a second threshold will be exceeded.
  • an anomaly detector in a test bench 3 can also Validate this detector before it is used on an onboard engine, which facilitates matters and reduces detector certification costs.
  • anomaly detectors can be used at low or at high input frequencies.
  • a first anomaly detector implements a normal behaviour model and makes a measurement (or score) of normality by a likelihood calculation.
  • This first detector uses low frequency time data segments as input data. It is used to detect abnormal behaviour of endogenic indicator vectors according to the context classes thus detecting abnormal operation of the test bench 3 and the engine component 5 .
  • a second anomaly detector implements a bearing damage detection model. It uses synchronous time segments of high frequency data (tachometers and accelerometers) that can be stored in files. After these files have been processed, the second detector generates a probability of an anomaly for each bearing and a detailed signature for each bearing.
  • a third anomaly detector implements an intermittent events detection model. This detector analyses vibration data acquired by segment, knowing that the response frequency is less than one response per segment. A warning message is sent when a confirmed intermittent event is detected. Another identification warning can be sent if the detection might lead to more precise information about the origin of the detected event.
  • FIG. 6 diagrammatically shows encapsulation of anomaly detectors in the monitoring system.
  • the monitoring system 1 comprises supervision means 61 to encapsulate anomaly detectors 51 , 52 , 53 and to distribute the data and monitor the behaviour of each anomaly detector.
  • These supervision means 61 comprise management means 63 , data distribution means 64 , parameter setting means 65 , communication means 66 and display means 67 .
  • the management means 63 are configured to start anomaly detectors 51 , 52 , 53 and to manage instances 71 , 72 , 73 corresponding to parameter setting and calibration choices of each of these anomaly detectors.
  • the data distribution means 64 are configured to procure and prepare input data for each of the anomaly detectors.
  • the parameter setting means 65 are configured to adjust instance parameters.
  • Communication means 66 are configured to receive, sort, select and read output messages from each anomaly detector.
  • the display means 67 are configured to display the results.
  • FIG. 7 diagrammatically shows a management system for a test bench of an engine component according to the invention.
  • the management system 81 comprises a control system 83 connected to the test bench 3 of an engine component 5 and a monitoring system 1 .
  • the control system 83 is configured to control the test bench 3 and to record data output from the test bench 3 and the engine component 5 in storage means 85 .
  • the monitoring system 1 is connected to the combination 11 of the test bench 3 and the engine component 5 through the control system 83 and through a data bus 87 that transmits data to it output from the combination 11 of the test bench 3 and the engine component 5 . This configuration is such that the functioning of the test bench is not disturbed.
  • control system 83 can be deployed in a first computer and the monitoring system 1 can be deployed in a second computer at a distance from the first.
  • the invention also relates to a computer program product that can be used in the different elements of the monitoring system, these programs comprising code instructions adapted to application of a method according to the invention as described above.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Chemical & Material Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Combustion & Propulsion (AREA)
  • Testing Of Engines (AREA)
  • Testing And Monitoring For Control Systems (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
US13/878,107 2010-10-11 2011-09-30 Monitoring system for an engine test bench Active 2035-02-12 US10359339B2 (en)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
FR1058236 2010-10-11
FR1058236A FR2965915B1 (fr) 2010-10-11 2010-10-11 Systeme de surveillance d'un banc d'essai de moteur d'aeronef
PCT/FR2011/052284 WO2012049396A1 (fr) 2010-10-11 2011-09-30 Système de surveillance d'un banc d'essai de moteur

Publications (2)

Publication Number Publication Date
US20130211768A1 US20130211768A1 (en) 2013-08-15
US10359339B2 true US10359339B2 (en) 2019-07-23

Family

ID=43983923

Family Applications (1)

Application Number Title Priority Date Filing Date
US13/878,107 Active 2035-02-12 US10359339B2 (en) 2010-10-11 2011-09-30 Monitoring system for an engine test bench

Country Status (9)

Country Link
US (1) US10359339B2 (pt)
EP (1) EP2627982B1 (pt)
JP (1) JP2013542432A (pt)
CN (1) CN103154693A (pt)
BR (1) BR112013008623A2 (pt)
CA (1) CA2813556C (pt)
FR (1) FR2965915B1 (pt)
RU (1) RU2013121587A (pt)
WO (1) WO2012049396A1 (pt)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220185502A1 (en) * 2019-04-01 2022-06-16 Safran Aircraft Engines Method for monitoring at least one aircraft engine

Families Citing this family (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2965915B1 (fr) 2010-10-11 2013-08-02 Snecma Systeme de surveillance d'un banc d'essai de moteur d'aeronef
FR2968038B1 (fr) 2010-11-26 2012-12-28 Snecma Systeme de detection d'un evenement fugace sur une roue aubagee de moteur d'aeronef
FR2986269B1 (fr) 2012-01-30 2015-08-07 Snecma Systeme de detection d'un impact sur une roue aubagee de moteur d'aeronef
FR2988130B1 (fr) 2012-03-13 2014-05-09 Snecma Systeme de detection de defaut sur une roue aubagee de moteur d'aeronef
FR2997451B1 (fr) 2012-10-26 2015-01-16 Snecma Systeme de surveillance d'un ensemble de composants moteur
FR3011936B1 (fr) 2013-10-11 2021-09-17 Snecma Procede, systeme et programme d'ordinateur d'analyse acoustique d'une machine
FR3011946B1 (fr) 2013-10-11 2016-07-08 Snecma Surveillance d'un moteur d'aeronef pour anticiper les operations de maintenance
FR3012930B1 (fr) * 2013-11-05 2015-12-25 Snecma Procede d'essai technique
FR3012882B1 (fr) * 2013-11-05 2015-11-27 Snecma Procede d'essai technique
FR3015670B1 (fr) 2013-12-20 2018-08-10 Safran Aircraft Engines Dispositif de detection de premices de defaillance d'un systeme mecanique
FR3019295B1 (fr) 2014-03-27 2016-03-18 Snecma Procede d'estimation du caractere normal ou non d'une valeur mesuree d'un parametre physique d'un moteur d'aeronef
FR3027667B1 (fr) 2014-10-22 2020-10-09 Snecma Procede et dispositif de surveillance d'une roue aubagee de moteur d'aeronef par mesure de position d'equilibre
FR3028067B1 (fr) 2014-11-05 2016-12-30 Snecma Outil de validation d'un systeme de surveillance d'un moteur d'aeronef
FR3028331B1 (fr) * 2014-11-10 2016-12-30 Snecma Procede de surveillance d'un moteur d'aeronef en fonctionnement dans un environnement donne
CN104807642B (zh) * 2015-03-31 2017-06-16 东软集团股份有限公司 汽车发动机故障检测方法和装置
CN104977169B (zh) * 2015-04-15 2017-12-22 北京宇航系统工程研究所 一种火箭发动机冷摆数字试验方法
FR3037170B1 (fr) 2015-06-03 2017-06-23 Snecma Procede et systeme de prediction du fonctionnement d'un aeronef par analyse de similarite utilisant des capacites de stockage et de calcul reparties
CN105865794B (zh) * 2016-05-12 2018-02-02 长安大学 基于短时傅立叶变换和主分量分析的发动机失火故障诊断方法
US10549862B1 (en) * 2018-09-05 2020-02-04 General Electric Company Method and system for smart and continuous engine operation monitoring
CN117235511B (zh) * 2023-11-13 2024-02-20 北京市计量检测科学研究院 一种二次仪表校准方法

Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB249971A (en) 1925-02-05 1926-04-08 Bohumil Jirotka Process for producing metal coatings on articles of aluminium and aluminium alloys
US4821217A (en) 1987-01-12 1989-04-11 The Boeing Company Programmable jet engine test station
EP1195593A1 (fr) 2000-10-05 2002-04-10 Techspace Aero S.A. Installation d'essais pour la réalisation de tests de moteurs d'avion
EP1281948A1 (fr) 2001-08-03 2003-02-05 Drecq Daniel Technologies D 2 T Programme informatique de contrôle et de commande pour banc d'essai
EP1715165A2 (de) 2005-04-21 2006-10-25 IAV GmbH Ingenieurgesellschaft Auto und Verkehr Verfahren und Vorrichtung zur Fehlerdiagnose für Verbrennungsmotoren
US7403850B1 (en) 2005-09-29 2008-07-22 Dynalco Controls Corporation Automated fault diagnosis method and system for engine-compressor sets
US20100063674A1 (en) 2008-09-11 2010-03-11 Assembly & Test Worldwide, Inc. Engine test method using structured test protocol
WO2012049396A1 (fr) 2010-10-11 2012-04-19 Snecma Système de surveillance d'un banc d'essai de moteur
FR2968038A1 (fr) 2010-11-26 2012-06-01 Snecma Systeme de detection d'un evenement fugace sur une roue aubagee de moteur d'aeronef
US20130197747A1 (en) 2012-01-30 2013-08-01 Snecma System for detecting an impact on an aircraft engine impeller wheel
US20130239653A1 (en) 2012-03-13 2013-09-19 Snecma System for detecting defects on an aircraft engine impeller wheel
WO2014064396A2 (fr) 2012-10-26 2014-05-01 Snecma Système de surveillance d'un ensemble de composants d'un équipement

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US1281948A (en) * 1917-07-28 1918-10-15 Kenneth S Guiterman Nursing-bottle holder.
CN2171854Y (zh) * 1993-09-08 1994-07-13 孙林 一种智能型发动机检测仪
JP3416396B2 (ja) * 1996-05-30 2003-06-16 三菱重工業株式会社 航空機エンジンテスト運転用設備のモニタ制御装置
JP2006292734A (ja) * 2005-03-15 2006-10-26 Omron Corp 検査装置用の判定モデル作成支援装置および検査装置ならびに耐久試験装置および耐久試験方法
DE602005008706D1 (de) * 2005-03-24 2008-09-18 Abb Research Ltd Abschätzen der Zustandparameter oder Erscheinung eines alternden Systems
JP5290026B2 (ja) * 2009-03-31 2013-09-18 日立建機株式会社 作業機械の学習診断システム、状態診断装置及び状態学習装置

Patent Citations (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB249971A (en) 1925-02-05 1926-04-08 Bohumil Jirotka Process for producing metal coatings on articles of aluminium and aluminium alloys
US4821217A (en) 1987-01-12 1989-04-11 The Boeing Company Programmable jet engine test station
EP1195593A1 (fr) 2000-10-05 2002-04-10 Techspace Aero S.A. Installation d'essais pour la réalisation de tests de moteurs d'avion
EP1281948A1 (fr) 2001-08-03 2003-02-05 Drecq Daniel Technologies D 2 T Programme informatique de contrôle et de commande pour banc d'essai
EP1715165A2 (de) 2005-04-21 2006-10-25 IAV GmbH Ingenieurgesellschaft Auto und Verkehr Verfahren und Vorrichtung zur Fehlerdiagnose für Verbrennungsmotoren
US7403850B1 (en) 2005-09-29 2008-07-22 Dynalco Controls Corporation Automated fault diagnosis method and system for engine-compressor sets
US20100063674A1 (en) 2008-09-11 2010-03-11 Assembly & Test Worldwide, Inc. Engine test method using structured test protocol
CN103154693A (zh) 2010-10-11 2013-06-12 斯奈克玛 发动机测试台的监控系统
WO2012049396A1 (fr) 2010-10-11 2012-04-19 Snecma Système de surveillance d'un banc d'essai de moteur
US20130211768A1 (en) 2010-10-11 2013-08-15 Snecma Monitoring system for an engine test bench
EP2627982A1 (fr) 2010-10-11 2013-08-21 Snecma Système de surveillance d'un banc d'essai de moteur
JP2013542432A (ja) 2010-10-11 2013-11-21 スネクマ エンジン試験台用監視システム
FR2968038A1 (fr) 2010-11-26 2012-06-01 Snecma Systeme de detection d'un evenement fugace sur une roue aubagee de moteur d'aeronef
US20120148400A1 (en) 2010-11-26 2012-06-14 Snecma System for detecting an ephemeral event on a vane impeller of an aircraft engine
US20130197747A1 (en) 2012-01-30 2013-08-01 Snecma System for detecting an impact on an aircraft engine impeller wheel
FR2986269A1 (fr) 2012-01-30 2013-08-02 Snecma Systeme de detection d'un impact sur une roue aubagee de moteur d'aeronef
US20130239653A1 (en) 2012-03-13 2013-09-19 Snecma System for detecting defects on an aircraft engine impeller wheel
FR2988130A1 (fr) 2012-03-13 2013-09-20 Snecma Systeme de detection de defaut sur une roue aubagee de moteur d'aeronef
WO2014064396A2 (fr) 2012-10-26 2014-05-01 Snecma Système de surveillance d'un ensemble de composants d'un équipement
FR2997451A1 (fr) 2012-10-26 2014-05-02 Snecma Systeme de surveillance d'un ensemble de composants moteur

Non-Patent Citations (11)

* Cited by examiner, † Cited by third party
Title
A. P. Dempster, et al., "Maximum Likelihood from Incomplete Data via the EM Algorithm", Journal of the Royal Statistical society, 39(1):1-38, 1977, 38 pages.
English machine translation of EP 1281948, retrieved Sep. 8, 2016. *
International Search Report dated Feb. 1, 2012 in PCT/FR2011/052284.
Jérôme Lacaille, "Standardized Failure Signature for a Turbofan Engine", IEEE Aerospace Conference, Big Sky, MT, 2009, 8 pages.
Jérôme Lacaille, "Validation of Health-Monitoring Algorithms for Civil Aircraft Engines", IEEE Aerospace Conference, Big Sky, MT, 2010, 11 pages.
U.S. Appl. No. 13/302,054, filed Nov. 22, 2011, 2012-0148400, Gerez et al.
U.S. Appl. No. 13/754,186, filed Jan. 30, 2013, 2013-0197747, Tourin et al.
U.S. Appl. No. 13/792,537, filed Mar. 11, 2013, 2013-0239653, Nicq et al.
U.S. Appl. No. 13/878,107, filed Apr. 5, 2013, 2013-0211768, Gerez, et al.
U.S. Appl. No. 14/438,129, filed Apr. 23, 2015, Lacaille, et al.
U.S. Appl. No. 14/575,096, filed Dec. 18, 2014, Gerez.

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220185502A1 (en) * 2019-04-01 2022-06-16 Safran Aircraft Engines Method for monitoring at least one aircraft engine
US12012231B2 (en) * 2019-04-01 2024-06-18 Safran Aircraft Engines Method for monitoring at least one aircraft engine

Also Published As

Publication number Publication date
US20130211768A1 (en) 2013-08-15
EP2627982B1 (fr) 2019-04-03
CN103154693A (zh) 2013-06-12
FR2965915B1 (fr) 2013-08-02
JP2013542432A (ja) 2013-11-21
FR2965915A1 (fr) 2012-04-13
CA2813556C (fr) 2019-01-15
CA2813556A1 (fr) 2012-04-19
BR112013008623A2 (pt) 2016-06-21
EP2627982A1 (fr) 2013-08-21
RU2013121587A (ru) 2014-11-20
WO2012049396A1 (fr) 2012-04-19

Similar Documents

Publication Publication Date Title
US10359339B2 (en) Monitoring system for an engine test bench
US7415328B2 (en) Hybrid model based fault detection and isolation system
US6898554B2 (en) Fault detection in a physical system
US7369932B2 (en) System and method for turbine engine fault detection using discrete event system modeling
US6999884B2 (en) Bearing anomaly detection and location
EP1297313B1 (en) Monitoring the health of a power plant
US9818242B2 (en) Gas turbine engine anomaly detections and fault identifications
US7280941B2 (en) Method and apparatus for in-situ detection and isolation of aircraft engine faults
EP2015186B1 (en) Diagnostic systems and methods for predictive condition monitoring
EP3410308B1 (en) Predictive analysis system and method for analyzing and detecting machine sensor failures
EP1630633A2 (en) System for gas turbine health monitoring data fusion
US10032322B2 (en) Validation tool for an aircraft engine monitoring system
US10551818B2 (en) Fault detection methods and systems
KR102545672B1 (ko) 기계고장 진단 방법 및 장치
Pecho et al. Vibration fault detection of fuel pump using recurrence quantification analysis
CN111555899A (zh) 告警规则配置方法、设备状态监测方法、装置和存储介质
US10345194B2 (en) Detection device for initiating failures of a mechanical system
CN115270896A (zh) 一种用于识别航空发动机主轴承松动故障的智能诊断方法
RU2688340C2 (ru) Способ вибродиагностирования газотурбинного двигателя
Ghiocel et al. A new probabilistic risk-based fault diagnosis procedure for gas turbine engine performance
Babbar et al. Fuzzy clustering based fault diagnosis for aircraft engine health management
Lacaille Robust monitoring of turbofan sensors
CN118333603A (zh) 一种飞机发动机egt状态监控方法、装置、设备及介质
CN118960824A (zh) 用于通过使用根据交通工具的机械系统的运行参数而变化的阈值来监测该系统的健康的方法和装置
Stramiello et al. Aviation Diagnostic and Engine Prognostic Technology (ADEPT) For the Chinook's T-55 Engine

Legal Events

Date Code Title Description
AS Assignment

Owner name: SNECMA, FRANCE

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:GEREZ, VALERIO;LACAILLE, JEROME HENRI NOEL;REEL/FRAME:030178/0217

Effective date: 20130301

AS Assignment

Owner name: SAFRAN AIRCRAFT ENGINES, FRANCE

Free format text: CHANGE OF NAME;ASSIGNOR:SNECMA;REEL/FRAME:046479/0807

Effective date: 20160803

AS Assignment

Owner name: SAFRAN AIRCRAFT ENGINES, FRANCE

Free format text: CORRECTIVE ASSIGNMENT TO CORRECT THE COVER SHEET TO REMOVE APPLICATION NOS. 10250419, 10786507, 10786409, 12416418, 12531115, 12996294, 12094637 12416422 PREVIOUSLY RECORDED ON REEL 046479 FRAME 0807. ASSIGNOR(S) HEREBY CONFIRMS THE CHANGE OF NAME;ASSIGNOR:SNECMA;REEL/FRAME:046939/0336

Effective date: 20160803

STPP Information on status: patent application and granting procedure in general

Free format text: NOTICE OF ALLOWANCE MAILED -- APPLICATION RECEIVED IN OFFICE OF PUBLICATIONS

STPP Information on status: patent application and granting procedure in general

Free format text: PUBLICATIONS -- ISSUE FEE PAYMENT RECEIVED

STPP Information on status: patent application and granting procedure in general

Free format text: PUBLICATIONS -- ISSUE FEE PAYMENT VERIFIED

STCF Information on status: patent grant

Free format text: PATENTED CASE

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1551); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Year of fee payment: 4